DreamVLA: A Vision-Language-Action Model Dreamed with Comprehensive World Knowledge

TL;DR

DreamVLA is a novel vision-language-action framework that integrates comprehensive world knowledge forecasting with dynamic, spatial, and semantic cues, enabling improved robot manipulation through a perception-prediction-action loop.

Contribution

It introduces a dynamic-region-guided knowledge prediction and a block-wise attention mechanism, along with a diffusion-based transformer, to enhance robot manipulation by better modeling world knowledge.

Findings

Achieves 76.7% success rate on real robot tasks

Attains 4.44 average length on CALVIN ABC-D benchmarks

Demonstrates improved generalization and reasoning in manipulation tasks

Abstract

Recent advances in vision-language-action (VLA) models have shown promise in integrating image generation with action prediction to improve generalization and reasoning in robot manipulation. However, existing methods are limited to challenging image-based forecasting, which suffers from redundant information and lacks comprehensive and critical world knowledge, including dynamic, spatial and semantic information. To address these limitations, we propose DreamVLA, a novel VLA framework that integrates comprehensive world knowledge forecasting to enable inverse dynamics modeling, thereby establishing a perception-prediction-action loop for manipulation tasks. Specifically, DreamVLA introduces a dynamic-region-guided world knowledge prediction, integrated with the spatial and semantic cues, which provide compact yet comprehensive representations for action planning. This design aligns with how humans interact with the world by first forming abstract multimodal reasoning chains before acting. To mitigate interference among the dynamic, spatial and semantic information during training, we adopt a block-wise structured attention mechanism that masks their mutual attention, preventing information leakage and keeping each representation clean and disentangled. Moreover, to model the conditional distribution over future actions, we employ a diffusion-based transformer that disentangles action representations from shared latent features. Extensive experiments on both real-world and simulation environments demonstrate that DreamVLA achieves 76.7% success rate on real robot tasks and 4.44 average length on the CALVIN ABC-D benchmarks.